What a Weld Record Has to Show Later
Why this matters
Nobody reads a weld record on the day it is written. It gets read months later, by somebody with a specific question: which joints came from a consumable lot that turned out to be wet, whether a welder was still current when a particular seam was run, which welds went in under the old revision of the procedure, or which other joints the same person made on the day one of them was rejected. Every one of those is a lookup, and a record answers a lookup only if it carries the field the question keys on. Six fields answer effectively all of them, and the resolution of the whole file is set by whichever one you left coarsest.
Where a record question turns into work: if the question arrived because a member has cracked in service, that is a live structural hazard and the sequence comes first, before anyone opens a folder: stop work under it, keep people out from under it, and get the member unloaded or independently supported by the party responsible for the structure. If the answer to a records question is that welds have to be removed and redone, the removal is its own exposure: gouging and grinding release the base metal and any coating as respirable dust, an inhalation route needing capture at the tool or a respirator issued under a written program meeting 29 CFR 1910.134, and air carbon arc gouging also puts the work inside a hearing conservation program under 29 CFR 1910.95.
The six fields, and the question each one answers
| Field | Written as | The question it answers later |
|---|---|---|
| Piece | The piece mark or assembly number that is physically marked on the part | "Which welds are on the assembly the customer is calling about?" |
| Joint | The joint identifier from the drawing or the weld map, not a description | "Which weld, out of the eighteen on that piece?" |
| Welder | The welder's stamp or assigned identifier | "Who deposited it, and what else did they run?" |
| Procedure | The procedure specification number and its revision | "Which welds predate the change we made to the procedure?" |
| Consumable | Classification plus lot or heat number | "Which welds used the lot that lost storage control?" |
| Date | The calendar date, and the shift if you run more than one | "Was the welder current, and what else went out that day?" |
Two of those fields are routinely written short and both cost the same way. A procedure number without its revision cannot separate work done before and after a change, which makes the revision field the one that decides whether a procedure change is a controlled event or an unrecoverable one. A consumable written as a classification without a lot identifies what kind of rod it was and never which rod it was, and the electrode storage card explains why the difference matters on a low-hydrogen product.
Resolution is set by the narrowest field
A weld record is a set of rows, and any question you ask returns the smallest group of rows the record can distinguish. Leave the joint field out and every question about one weld returns the whole assembly. Record the date but not the shift and every question about one shift returns a day. Record the welder at crew level rather than by individual and every question about one person returns the crew.
That is not a paperwork nicety, it is the size of the corrective action. When a joint is rejected and the governing document requires the examination to be expanded, the expansion is scoped by what the record can isolate. A record that can isolate the welder, the shift and the consumable lot expands the examination to a handful of joints. A record that can only isolate the assembly expands it to all of them.
What a weld record is not
It is not the welding procedure specification, and it is not the welder qualification record. Those are three documents about three different subjects, and the card on reading a procedure specification owns that distinction. The weld record is the only one of the three that says what actually happened on a specific length of weld.
It is also not an inspection result, though the two get filed together. An inspection result records what was examined, by which method, at what coverage and against which criterion. The weld record records what was deposited. Merging them into one line reliably loses one of the two, and the case card on the weld that passed visual and failed in service is what that looks like from the far end.
The artifact: one assembly's weld log
Assembly F-12, 18 joints, logged one row per joint. Four rows shown, in the format the whole log uses.
| Joint | Piece | Welder | WPS and rev | Consumable and lot | Date |
|---|---|---|---|---|---|
| F12-01 | F-12 | C | WPS-04 rev B | E71T-1 lot 4471 | 14 Mar |
| F12-07 | F-12 | C | WPS-04 rev B | E71T-1 lot 4471 | 14 Mar |
| F12-11 | F-12 | C | WPS-04 rev B | E71T-1 lot 4488 | 21 Mar |
| F12-18 | F-12 | B | WPS-04 rev C | E71T-1 lot 4488 | 02 Apr |
Two things are visible in four rows that a narrative note would have lost: the procedure revision changed between March and April, and the consumable lot changed between the 14th and the 21st. Neither is remarkable on the day. Both are the answer to a question somebody asks later.
Three questions run against it
Question 1, from the consumable side. A carton of lot 4471 was found outside storage control. Lookup on the consumable field returns 5 of the 18 joints on this assembly. Resolvable, and the corrective scope is 5 joints rather than 18. The electrode storage card owns what conditioning can and cannot restore; the record's job was only to hand over the list.
Question 2, from the qualification side. Is welder B still current for this process? Lookup on welder plus date returns B's most recent entry on this log as 02 Apr, four months back from the question. Under AWS D1.1, in the edition invoked by your project specification or the adopted building code, a welder's qualification remains in effect for a process unless that process has gone unused for a period exceeding six months. Four months is inside that, so on this evidence B is current, and the welder qualification card owns the continuity rule and its exceptions. Note what made this answerable: the date field, not the welder field.
Question 3, from the inspection side, and this is the one that does not fully resolve. Joint F12-11 was rejected. The governing document requires the examination to be expanded to the other work from the same shift. Which joints went out on the shift that produced it? The log carries a date but not a shift, so the lookup can only return every joint the shop ran on 21 Mar, which was 6, when the shop actually ran two shifts and 3 of those 6 belong to the shift F12-11 came from. The examination expansion therefore covers 6 joints instead of 3, because the record could not distinguish the shift.
That is the narrowest-field rule with a number on it: one missing column doubled the scope of the corrective action on this assembly, and it will do so on every assembly until the column is added.
What the log lets you compute, and what it does not
It lets you compute a repair rate, which is the one quality number a small shop can actually maintain. Rejected joints divided by joints examined, over a stated period.
- On this assembly: 1 rejection in 18 joints, which is 5.6 percent.
- Across the shop for the quarter: 4 rejections in 214 joints examined, which is 1.9 percent.
- Welder C over the quarter: 1 rejection in 22 joints, which is 4.5 percent.
Correct the comparator before you draw the conclusion. Comparing C's 4.5 percent against the shop's 1.9 percent is comparing C against a blend that already contains C. Take C out: the rest of the shop is 3 rejections in 192 joints, which is 1.6 percent. That is the honest comparator, and it makes C's figure look worse, not better, which is exactly why the correction has to be made before anybody uses the number in a conversation.
Then refuse to conclude from it anyway. C's rate rests on 22 joints and one event. One more rejection would put C at 2 in 22, or 9.1 percent; one fewer would put C at zero. A metric that swings between 0 and 9.1 percent on a single event is not measuring a person, and using it as though it were is how a shop loses a good welder. What the number is genuinely good for is watching the shop's own rate over time against itself, and for telling you when a rate moved enough to go looking for a cause in the other five fields.
It does not let you compute anything about the welds nobody examined. The 214 is joints examined, not joints made, and those are different denominators. Quoting a repair rate without saying which one it is on is the fastest way to make a real number useless.
Check the artifact against the rules the siblings state
- Count re-derived from the printed list, last mention first. The article names six fields and the table enumerates six rows: piece, joint, welder, procedure with revision, consumable with lot, and date. No later sentence refers to five or seven.
- Arithmetic re-derived. 1 over 18 is 5.56 percent, stated as 5.6. 4 over 214 is 1.87 percent, stated as 1.9. 1 over 22 is 4.55 percent, stated as 4.5. 3 over 192 is 1.56 percent, stated as 1.6. 2 over 22 is 9.09 percent, stated as 9.1.
- The comparator was corrected and the correction is printed. 4 rejections in 214 becomes 3 in 192 once C's 1 in 22 is taken out, and 214 minus 22 is 192 while 4 minus 1 is 3.
- The denominators are named and not swapped. 214 is joints examined for the quarter, 18 is joints on this assembly, 22 is C's joints in the quarter. The closing paragraph states plainly that joints examined is not joints made.
- Question 3 is reported as a partial answer, not a clean one. 6 joints returned where 3 was the true set, with the missing shift column named as the cause. The article does not narrate it as resolved.
- Direction check against the siblings. The welder qualification card owns continuity and the case card on the weld that failed in service owns what a scoped acceptance is worth. This card supplies the fields both of them look up, and cites rather than restates each rule.
References
- AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the building code your authority having jurisdiction has adopted, for weld identification, welder identification and qualification continuity
- ASME Boiler and Pressure Vessel Code Section IX, as invoked by the applicable piping or vessel code and adopted by your jurisdiction, where the work is pressure-retaining
- 29 CFR 1910.134 for respiratory protection and 29 CFR 1910.95 for hearing conservation, where a records question leads to removal and re-welding
- See related: How to Read a Welding Procedure Specification; What a Welder Qualification Covers and What It Does Not; What Electrode Storage Has to Do With Cracking; The Weld That Passed Visual and Failed in Service